What's under the hood ?
NVIDIA have been taking positive steps to move from the "fill rates" and "speed" to more of a cinematic computing. This movement from pipelines and fillrate has mean't a reduction from 8X pipelines to 4X pipelines. The core architecture of the NV30 or FX series of cards is based on purley pixel and vertext shaders - so instead of having a high fillrate you have more programability and optimisations avaliable to the developer.
Further, the FX5200 / Ultra incarnations, being budget cards, need to lose some of the features found on the more expensive models. Having lower clocked GPU and memory timings help in cost-cutting measures, but reducing total transistor count is more important. The FX5600 Ultra weighs in at around 80 million transistors; still far short of the 100 million+ offerings from the FX5800 Ultra and Radeon 9500/9700 cards. The FX5200, though, is a relative lightweight with 'only' ~ 45 million under the cooler. We're missing around 35 million transistors - where did they go ?. For starters, NVIDIA's 5200 line doesn't feature the same colour or z-compression found on the more expensive models. The former, as you may know, kind of limits the cards' performance in situations where mutli-sampling anti-aliasing is used. The memory-hogging process can't be helped on its way with the lower-priced cards. That may account for some of the transistor deficiencies in the cheaper 5200-based cards, but there may be more differences than this alone.
If these cards have the possibility of being slower than present NVIDIA cards, then why are they being released ?. Cinematic rendering is a term that has been bandied about of late. With DirectX9's API better able to take advantage of pixel and vertex shading, both the 5200 and 5600 cards feature DX9-busting features. Adhering to the API's Vertex Shader 2.0 specifications, and going beyond the Pixel Shader 2.0, these newer cards have the ability to showcase some pretty special effects. It's now just a question of games coders catching up.
Faster cores, better hardware features, and relatively low street prices should ensure that the cards mount a challenge to ATi's line-up. Less pixel-pushing power on an equal GPU speed basis when compared with the current Ti 4xxx line, these cards hit back with grotesque core speeds and super-fast memory. An interesting change. It seems as if the shift from pure fill-rate to concentrating on advanced features has begun in earnest. Let's see how they fare as we benchmark them against some of their immediate competition. (T.S)
